ABSTRACT
Photochemical and advanced oxidation processes (AOPs) are promising options to simultaneously disinfect wastewater treatment plant (WWTP) effluent and degrade organic micropollutants (OMPs). In the present study, kinetic experiments with UV alone (254 or 254+185 nm) and the combination of UV and hydrogen peroxide (H2O2) as an AOP were conducted in WWTP effluent (with and without ultrafiltration) to assess the degradation of 24 OMPs, 13 of which were examined for the first time, in a flow-through batch reactor setup. Four parameters were systematically varied to quantify their impacts on OMP degradation: H2O2 concentrations, UV source, water matrix, and flow rate. Most of the studied OMPs (e.g., trimethylammonium, 1,3-di-o-tolylguanidine, carbamazepine) were only significantly degraded in the presence of UV radiation and H2O2. The highest pseudo-first-order rate constants were found for diclofenac, acesulfame, diatrizoate, and sulfamethoxazole. The experiment with the highest degradation also showed the strongest abatement of UV absorbance at 254 nm. Only three of the 24 substances (cyanoguanidine, melamine, and oxipurinol) were not degraded; UV alone even increased their concentrations. Overall, upgrading a UV disinfection to an AOP using H2O2 allows the degradation of a wide range of OMPs, making it an interesting process for water reuse.
HIGHLIGHTS
The UV/H2O2 process could be a promising advanced oxidation process for water reuse.
The degradation of 24 organic micropollutants (OMPs) was studied, 13 of which were examined for the first time.
UV/H2O2 was able to degrade 21 of these OMPs.
A design of experiments approach was used to model degradation as a function of the process settings.
The resulting regression equations allow substance-specific optimization of the process.
INTRODUCTION
With water scarcity and climate variability increasingly affecting water supplies, the importance of water reuse in reducing water stress continues to grow (Asano et al. 2007; Tram Vo et al. 2014; Drechsel et al. 2022). Water reuse has applications across various sectors, including agriculture, urban environments, industry, leisure activities, water body restoration, groundwater recharge, and the production of potable water (Salgot & Folch 2018; Drechsel et al. 2022). Several treatment steps are typically required before reuse, with UV treatment often applied for disinfection after physical, biological, and chemical treatments (Tram Vo et al. 2014; Jeffrey et al. 2022). UV treatment can inactivate a range of pathogens by damaging their nucleic acids (Asano et al. 2007; Koutchma 2019). UV irradiation of wastewater for disinfection can additionally contribute to the degradation of organic micropollutants (OMPs) through direct photolysis (Klöpffer 2012).
The occurrence of OMPs in the environment, particularly of persistent and mobile (PM) organic substances with unknown toxicological properties, poses risks to both the environment and human health (Reemtsma et al. 2016; Neuwald et al. 2021). PM substances accumulate in the water cycle due to their high solubility and low sorption (Hembrock-Heger 2007; Reemtsma et al. 2016; Zeeshan et al. 2023b). Furthermore, some OMPs accumulate in humans, plants, and soil (Reemtsma et al. 2016; Arp et al. 2017; Zahn et al. 2020; Dittmann et al. 2024). Some OMPs are highly susceptible to photolysis, including diclofenac (DCF) and sulfamethoxazole (SMX), whereas others are not susceptible to photolysis but have been shown to be degraded in UV advanced oxidation processes (AOPs), where UV light is combined with the addition of chemical oxidants such as H2O2, for example, carbamazepine (CBZ) and primidone (PRI) (Yu et al. 2015; Miklos et al. 2018).
UV radiation combined with hydrogen peroxide (UV/H2O2) as an AOP generates hydroxyl radicals that degrade OMPs through nonspecific oxidation with higher reaction rates compared to UV radiation alone (Wols et al. 2013; Mierzwa et al. 2018). When applying light energy at a wavelength around 254 nm, hydroxyl radicals are formed through the homolytic cleavage of H2O2 molecules (Mierzwa et al. 2018). When using a wavelength of 185 nm, water is directly split into hydroxyl radicals (Zoschke et al. 2014; Koutchma 2019). The degradation of OMPs by AOP involving H2O2 has been investigated in several studies (Lopez et al. 2003; Rosario-Ortiz et al. 2010; Bolton et al. 2015; Wols et al.,2013, 2015; Yu et al. 2015). Some OMPs, such as DCF, SMX, and CBZ, were frequently studied using collimated beam experiments to quantify degradation with kinetic parameters such as reaction constants. Previous studies have already modeled OMP degradation by UV/H2O2-AOP. For example, Wols et al. (2015) and Yu et al. (2015) revealed that the water matrix and temperature have an important impact on OMP degradation. A computational fluid dynamics model was created in a study by Wols et al. (2015), enhancing the UV/H2O2–AOP model with insights into particle tracks and the UV dose distribution. Fluorescence spectroscopy has been used to further characterize the UV/H2O2–AOP (Yu et al. 2015), exhibiting stronger abatement of fluorescence during AOP treatment in the presence of H2O2 compared to solely UV.
Photolytic transformation of pharmaceuticals can yield byproducts, some of which have been found to exhibit toxic properties on fish, daphnia, and algae (Wang et al. 2012; Yan & Song 2014). UV-AOP processes are well known for degrading particular OMP, but it is important to note that the exact transformation mechanisms for many wastewater chemicals remain unknown. UV absorbance at 254 nm (UVA254) has been reported to be a useful surrogate for OMP elimination by oxidation with ozone (Altmann et al. 2016), adsorption onto activated carbon (Zietzschmann et al. 2014), and in UV/H2O2–AOP experiments with collimated beams (Rosario-Ortiz et al. 2010; Yu et al. 2015).
The eliminations of recently detected PM OMPs by UV/H2O2–AOP in wastewater treatment plant (WWTP) effluents are poorly known. Therefore, this study aimed to investigate the removal of 24 potentially PM OMPs in WWTP effluent by the AOP (UV/H2O2 processes) in a flow-through batch reactor setup. These substances represent inputs from a range of origins, including health care, industry, and agriculture. A total of 14 of the substances have been detected in surface and groundwater (Arp & Hale 2019; Schulze et al. 2019; Zahn et al. 2019; Huang et al. 2021; Liu et al. 2021; Neuwald et al. 2022; Scheurer et al. 2022), and 9 of the substances were found to be persistent in river bank filtration (Zeeshan et al. 2023b), underlining the importance of studying their removal from WWTP effluents by UV and UV/H2O2–AOP. The degradation of 13 of these OMPs with UV/H2O2 is studied for the first time. Using a design of experiment approach, this study systematically varies four of the main process parameters (H2O2 concentration, UV source, water matrix, and flow rate) and assesses the effect on OMP degradation and reaction rate constants. To assess the potential to monitor the UV/H2O2 process, this study additionally investigates the impact of the process on the dissolved organic matter (DOM) composition and compares degradation constants with the abatement of UVA254.
MATERIALS AND METHODS
Organic micropollutants
The 24 investigated OMPs are listed in Table 1. A mixed stock solution with a concentration of 10 mg/L was prepared by dissolving 10 mg of each substance in 1 L of ultrapure water.
Substance list of the 24 OMPs
OMP . | Abbreviation . | CAS No. . | Mean retention time (min) . | Literature on H2O2–AOP treatment . |
---|---|---|---|---|
2-Acrylamino-2-methylpropane sulfonate | AAMPS | 5165-97-9 | 8.19 | n.a. |
Acesulfame | ACE | 55589-62-3 | 7.11 | Wang et al. (2019); Wünsch et al. 2019 |
Adamantan-1-amine | ATA | 768-94-5 | 9.05 | n.a. |
Benzyldimethylamine | BDMA | 103-83-3 | 6.58 | n.a. |
Benzyl-trimethylammonium | BETMAC | 56-93-9 | 6.57 | n.a. |
Benzotriazole | BTA | 95-14-7 | 9.93 | Bahnmüller et al. (2015); Yu et al. (2015); Miklos et al. (2019, 2018); Dang et al. (2024) |
Carbamazepine | CBZ | 298-46-4 | 13.02 | Hembrock-Heger (2007); Kim et al. (2009); Wols & Hofman-Caris (2012); Wols et al. (2013); Yu et al. (2015); Afonso-Olivares et al. (2016); Miklos et al. (2019, 2018) |
Cyanoguanidine | CG | 461-58-5 | 1.57 | n.a. |
Diclofenac | DCF | 15307-86-5 | 15.21 | Kim et al. (2009); Wols & Hofman-Caris (2012); Wols et al. (2013); Yu et al. (2015); Afonso-Olivares et al. (2016); Miklos et al. (2018); Becerril et al. (2019) |
1,3-Di-o-tolylguanidine | DIOTOG | 97-39-2 | 9.73 | n.a. |
Dimethylbenzenesulfonate | DMBSA/XSA | 25321-41-9 1300-72-7 | 11.78 | Mazellier et al. (2004); Kalinski et al. (2014) |
1,3-Diphenylguanidine | DPG | 102-06-7 | 8.96 | n.a. |
Diatrizoate | DZA | 117-96-4 | 7.95 | Wols & Hofman-Caris (2012); Wols et al. (2013); Velo-Gala et al. (2014); Duan et al. (2017) |
4-Hydroxy-1-(2hydroxyehtyl)-2,2,6,6, tetramethylpiperidine | HHTMP | 52722-86-8 | 2.77 | n.a. |
N-(3-(dimethylamino)propyl) methacrylamide | MAPMA | 5205-93-6 | 4.54 | Chen et al. (2011) |
Melamine | MEL | 108-78-1 | 1.29 | n.a. |
2-Methyl-2-propene-1-sulfonate | MPSA | 1561-92-8 | 6.01 | n.a. |
Oxipurinol | OXP | 2465-59-0 | 5.43 | n.a. |
Primidon | PRI | 125-33-7 | 10.82 | Wols & Hofman-Caris (2012); Yu et al. (2015); Nihemaiti et al. (2018); Miklos et al. (2019, 2018) |
p-Toluenesulfonate | PTSS | 104-15-4 | 9.88 | n.a. |
Saccharine | SAC | 81-07-2 | 8.96 | Ye et al. (2022) |
Sulfamethoxazole | SMX | 723-46-6 | 9.99 | Hembrock-Heger (2007); Kim et al. (2009); Wols & Hofman-Caris (2012); Wols et al. (2013); Yu et al. (2015); Afonso-Olivares et al. (2016); Miklos et al. (2018) |
Trifluoromethanesulfonate | TFMSA | 1493-13-6 | 3.59 | n.a. |
Valsartanate | VSA | 164265-78-5 | 12.41 | Martínez-Pachón et al. (2019); Castro et al. (2021) |
OMP . | Abbreviation . | CAS No. . | Mean retention time (min) . | Literature on H2O2–AOP treatment . |
---|---|---|---|---|
2-Acrylamino-2-methylpropane sulfonate | AAMPS | 5165-97-9 | 8.19 | n.a. |
Acesulfame | ACE | 55589-62-3 | 7.11 | Wang et al. (2019); Wünsch et al. 2019 |
Adamantan-1-amine | ATA | 768-94-5 | 9.05 | n.a. |
Benzyldimethylamine | BDMA | 103-83-3 | 6.58 | n.a. |
Benzyl-trimethylammonium | BETMAC | 56-93-9 | 6.57 | n.a. |
Benzotriazole | BTA | 95-14-7 | 9.93 | Bahnmüller et al. (2015); Yu et al. (2015); Miklos et al. (2019, 2018); Dang et al. (2024) |
Carbamazepine | CBZ | 298-46-4 | 13.02 | Hembrock-Heger (2007); Kim et al. (2009); Wols & Hofman-Caris (2012); Wols et al. (2013); Yu et al. (2015); Afonso-Olivares et al. (2016); Miklos et al. (2019, 2018) |
Cyanoguanidine | CG | 461-58-5 | 1.57 | n.a. |
Diclofenac | DCF | 15307-86-5 | 15.21 | Kim et al. (2009); Wols & Hofman-Caris (2012); Wols et al. (2013); Yu et al. (2015); Afonso-Olivares et al. (2016); Miklos et al. (2018); Becerril et al. (2019) |
1,3-Di-o-tolylguanidine | DIOTOG | 97-39-2 | 9.73 | n.a. |
Dimethylbenzenesulfonate | DMBSA/XSA | 25321-41-9 1300-72-7 | 11.78 | Mazellier et al. (2004); Kalinski et al. (2014) |
1,3-Diphenylguanidine | DPG | 102-06-7 | 8.96 | n.a. |
Diatrizoate | DZA | 117-96-4 | 7.95 | Wols & Hofman-Caris (2012); Wols et al. (2013); Velo-Gala et al. (2014); Duan et al. (2017) |
4-Hydroxy-1-(2hydroxyehtyl)-2,2,6,6, tetramethylpiperidine | HHTMP | 52722-86-8 | 2.77 | n.a. |
N-(3-(dimethylamino)propyl) methacrylamide | MAPMA | 5205-93-6 | 4.54 | Chen et al. (2011) |
Melamine | MEL | 108-78-1 | 1.29 | n.a. |
2-Methyl-2-propene-1-sulfonate | MPSA | 1561-92-8 | 6.01 | n.a. |
Oxipurinol | OXP | 2465-59-0 | 5.43 | n.a. |
Primidon | PRI | 125-33-7 | 10.82 | Wols & Hofman-Caris (2012); Yu et al. (2015); Nihemaiti et al. (2018); Miklos et al. (2019, 2018) |
p-Toluenesulfonate | PTSS | 104-15-4 | 9.88 | n.a. |
Saccharine | SAC | 81-07-2 | 8.96 | Ye et al. (2022) |
Sulfamethoxazole | SMX | 723-46-6 | 9.99 | Hembrock-Heger (2007); Kim et al. (2009); Wols & Hofman-Caris (2012); Wols et al. (2013); Yu et al. (2015); Afonso-Olivares et al. (2016); Miklos et al. (2018) |
Trifluoromethanesulfonate | TFMSA | 1493-13-6 | 3.59 | n.a. |
Valsartanate | VSA | 164265-78-5 | 12.41 | Martínez-Pachón et al. (2019); Castro et al. (2021) |
Note. DMBSA and XSA are positional isomers and are detected together at the same retention time. Substances for which the degradation with AOPs has been reported in the literature are highlighted in bold. n.a.: not available.
Experimental setup
Scheme of the bench-scale testing stand. Water was pumped from the mixing vessel through the UV reactor holding a UV lamp emitting light at 254 nm or at 254 + 185 nm. H2O2 was added at the beginning of some of the experiments. During each experiment, seven samples were taken from the mixing vessel for OMP analysis. The reactor volume is situated between the quartz glass and the steel case, as depicted with blue circles in the cross-section of the reactor. The black circle represents the lamp itself. A photograph of the testing stand can be found in the extensive SI A.
Scheme of the bench-scale testing stand. Water was pumped from the mixing vessel through the UV reactor holding a UV lamp emitting light at 254 nm or at 254 + 185 nm. H2O2 was added at the beginning of some of the experiments. During each experiment, seven samples were taken from the mixing vessel for OMP analysis. The reactor volume is situated between the quartz glass and the steel case, as depicted with blue circles in the cross-section of the reactor. The black circle represents the lamp itself. A photograph of the testing stand can be found in the extensive SI A.
In order to reach a minimum of 1 μg/L of each OMP, the stock solution was spiked into the mixing vessel. The water in the mixing vessel was recirculated by a pump with flow rates of 40, 76, or 100 mL/s. As specified in Table 2, two separate low-pressure mercury radiation sources (UV-EL, Germany) with 5.4 W irradiation power in the UV-C range were tested. The lamps were either a monochromatic UV lamp with emission at only 254 nm wavelength or a dichromatic UV lamp with around 20% additional irradiation intensity at 185 nm. About 30% H2O2 solution (Merck-Schuchardt, Germany) was used to reach initial H2O2 concentrations of 0, 10, or 20 mg/L in the mixing vessel.
Conducted experiments according to the combination of four factors (DoE) with central points marked with bold numbers
Experiment number . | H2O2 concentration (mg/L) . | Water matrix . | UV lamp (nm) . | Flow rate (mL/s) . |
---|---|---|---|---|
1 | 20 | UF permeate | 254 | 100 |
2 | 0 | WWTP effluent | 254 | 100 |
3 | 10 | Mix | 254 + 185 | 76 |
4 | 10 | Mix | 254 + 185 | 76 |
5 | 20 | UF permeate | 254 + 185 | 40 |
6 | 0 | UF permeate | 254 | 40 |
7 | 10 | Mix | 254 | 76 |
8 | 0 | UF permeate | 254 + 185 | 100 |
9 | 20 | WWTP effluent | 254 | 40 |
10 | 10 | Mix | 254 + 185 | 76 |
11 | 0 | WWTP effluent | 254 + 185 | 40 |
12 | 10 | Mix | 254 | 76 |
13 | 10 | Mix | 254 | 76 |
14 | 20 | WWTP effluent | 254 + 185 | 100 |
15 | 0 | WWTP effluent | Without | 100 |
16 | 0 | RO | Without | 100 |
Experiment number . | H2O2 concentration (mg/L) . | Water matrix . | UV lamp (nm) . | Flow rate (mL/s) . |
---|---|---|---|---|
1 | 20 | UF permeate | 254 | 100 |
2 | 0 | WWTP effluent | 254 | 100 |
3 | 10 | Mix | 254 + 185 | 76 |
4 | 10 | Mix | 254 + 185 | 76 |
5 | 20 | UF permeate | 254 + 185 | 40 |
6 | 0 | UF permeate | 254 | 40 |
7 | 10 | Mix | 254 | 76 |
8 | 0 | UF permeate | 254 + 185 | 100 |
9 | 20 | WWTP effluent | 254 | 40 |
10 | 10 | Mix | 254 + 185 | 76 |
11 | 0 | WWTP effluent | 254 + 185 | 40 |
12 | 10 | Mix | 254 | 76 |
13 | 10 | Mix | 254 | 76 |
14 | 20 | WWTP effluent | 254 + 185 | 100 |
15 | 0 | WWTP effluent | Without | 100 |
16 | 0 | RO | Without | 100 |
Note. Additional experiments 15 and 16 were conducted for quality control (additional information is provided in SI D).
The duration of each experiment was 3 h, with sampling before the start and after 10, 20, 40, 60, 120, and 180 min. Preliminary experiments showed no OMP removal by H2O2 alone (at 50 mg/L) in the water matrix (UF permeate); therefore, no addition of quenching agents was needed to stop H2O2 reactions after sampling (Supporting Information (SI) B), which is consistent with results from Miklos et al. (2018) for H2O2 concentrations of up to 20 mg/L.
Preparation of the water matrices
Experiments were conducted with three different water matrices (WWTP effluent, permeate from ultrafiltration of WWTP effluent, and a 1:1 mixture of both) to assess the influence of different water qualities on OMP degradation. Several grab samples of WWTP effluent were collected between April and August 2022 from a WWTP in Berlin, Germany (approximately 105,000 m3/day dry weather capacity, 10.3 mg/L of dissolved organic carbon (DOC)). The WWTP operates with primary settling followed by activated sludge treatment with biological phosphorus removal, denitrification and nitrification, and secondary clarification. WWTP effluent was then filtered with an ultrafiltration unit (LSta80-SPS, SimaTec, Germany) through a membrane with a nominal molecular weight cutoff of 150 kDa (Nadir UP150P, Mann + Hummel, Germany). Further information on the ultrafiltration procedure can be found in SI C.
Design of experiment
Parameter combinations of the conducted experiments are listed in Table 2 based on the design of experiment (DoE) approach using a half fractional factorial design (Resolution IV, one Block) in DoE software (Minitab statistical software, Minitab, USA). The number of experiments was thereby reduced from 54 (full factorial) to 14. Thus, the DoE approach allowed for assessing the individual effect of each factor on OMP degradation and linearly modeling OMP removals for all 54 possible combinations if the effect is linear. Non-linearity can be found with the help of center points, which describe the level in the middle of the lower and upper factor levels. Additionally, two quality control experiments were carried out with reverse osmosis permeate (RO) and WWTP effluent in order to check for degradation of OMPs without UV light.
Analytics
Samples were filtered (0.45 μm, regenerated cellulose) and stored in a fridge at 3 °C immediately after sampling. Temperature, pH (SenTix, Xylem WTW, USA), and conductivity (TetraCon, Xylem WTW, USA) were measured using pH and conductivity probes, respectively.
UVA254 was measured using a spectrophotometer (Lambda 25, Perkin-Elmer, USA), while DOC and total organic carbon (TOC) concentrations were quantified with a TOC analyzer (varioTOC cube, Elementar, Germany). DOC and TOC samples (each 5 mL) were diluted 1:1 with deionized water from Milli-Q IQ7000 with Millipak 0.22 μm filter (Merck, Germany).
The DOC compositions of selected samples were further characterized with fluorescence spectroscopy (FluoroMax-4, Horiba, Japan) according to Zeeshan et al. (2023a). In addition, size-exclusion chromatography (HW-50S column, Alltech-GROM, Germany) was employed, coupled with continuous UV analyses (liquid chromatography UV detection, LC-UVD) and DOC quantification liquid chromatography organic carbon detection, (LC-OCD, DOC-Labor Dr Huber, Germany) following the methodologies established by Haberkamp et al. (2007) and Huber et al. (2011). LC-OCD-UVD data were evaluated using the software LC-OCD converter (Saal 2022) and Origin (OriginLab, USA). LC-OCD-UVD samples were diluted 1:1 with deionized water.
OMP concentrations were measured using a high-performance liquid chromatography system (Agilent 1290 Infinity, Agilent, USA) with a 5,500 triple-quadrupole mass spectrometer (MS/MS) with Turbo V ion source (Sciex, USA) and a reverse phase column (Atlantis T3, Waters, USA) according to a method described by Zeeshan et al. (2023b).
Modelling
Degradation constants (d-values)
The degradation constant is the slope of the logarithmically linearized concentration plots. The logarithmically linearized concentration plots are calculated using Equation (2), where d is the degradation constant (L/min), c0 is the initial concentration of one specific OMP (μg/L) at the beginning of the experiment (time t0 = 0) and c is the concentration (μg/L) at a certain time t (min). The plots were created automatically using a code written in R (R Core Team 2022). More information can be found in SI E and the original data and code to create Table 3 and Figures 4–6 can be found online (https://doi.org/10.5281/zenodo.14051870). To account for fouling on the quartz glass, the time axis was corrected for each experiment, thus reducing the effective exposure time (further information can be found in SI F).
Degradation constants (d-values), pseudo-first-order rate constants (k-values), and removals after a single passage through the reactor are shown for all experiments for DIOTOG
Experiment number . | Flow rate . | Wavelength (nm) . | Water matrix . | H2O2 concentration (mg/L) . | Removal (-) (%) . | d (1/s) . | k (1/s) . |
---|---|---|---|---|---|---|---|
1 | High | 254 | UF permeate | 20 | 8 | 0.0008 | 0.02 |
14 | High | 254 + 185 | WWTP effluent | 20 | 8 | 0.0008 | 0.03 |
5 | Low | 254 + 185 | UF permeate | 20 | 14 | 0.0006 | 0.01 |
9 | Low | 254 | WWTP effluent | 20 | 14 | 0.0005 | 0.01 |
3 | Medium | 254 + 185 | Mix | 10 | 6 | 0.0004 | 0.01 |
4 | Medium | 254 + 185 | Mix | 10 | 6 | 0.0004 | 0.01 |
12 | Medium | 254 | Mix | 10 | 5 | 0.0004 | 0.01 |
7 | Medium | 254 | Mix | 10 | 5 | 0.0004 | 0.001 |
13 | medium | 254 | Mix | 10 | 5 | 0.0004 | 0.009 |
10 | Medium | 254 + 185 | Mix | 10 | 5 | 0.0004 | 0.009 |
8 | High | 254 + 185 | UF permeate | 0 | 2 | 0.0002 | 0.004 |
11 | Low | 254 + 185 | WWTP effluent | 0 | 2 | 0.0001 | 0.002 |
6 | Low | 254 | UF permeate | 0 | 1 | 0.0001 | 0.001 |
2 | High | 254 | WWTP effluent | 0 | 0 | 0.0000 | 0.0005 |
Experiment number . | Flow rate . | Wavelength (nm) . | Water matrix . | H2O2 concentration (mg/L) . | Removal (-) (%) . | d (1/s) . | k (1/s) . |
---|---|---|---|---|---|---|---|
1 | High | 254 | UF permeate | 20 | 8 | 0.0008 | 0.02 |
14 | High | 254 + 185 | WWTP effluent | 20 | 8 | 0.0008 | 0.03 |
5 | Low | 254 + 185 | UF permeate | 20 | 14 | 0.0006 | 0.01 |
9 | Low | 254 | WWTP effluent | 20 | 14 | 0.0005 | 0.01 |
3 | Medium | 254 + 185 | Mix | 10 | 6 | 0.0004 | 0.01 |
4 | Medium | 254 + 185 | Mix | 10 | 6 | 0.0004 | 0.01 |
12 | Medium | 254 | Mix | 10 | 5 | 0.0004 | 0.01 |
7 | Medium | 254 | Mix | 10 | 5 | 0.0004 | 0.001 |
13 | medium | 254 | Mix | 10 | 5 | 0.0004 | 0.009 |
10 | Medium | 254 + 185 | Mix | 10 | 5 | 0.0004 | 0.009 |
8 | High | 254 + 185 | UF permeate | 0 | 2 | 0.0002 | 0.004 |
11 | Low | 254 + 185 | WWTP effluent | 0 | 2 | 0.0001 | 0.002 |
6 | Low | 254 | UF permeate | 0 | 1 | 0.0001 | 0.001 |
2 | High | 254 | WWTP effluent | 0 | 0 | 0.0000 | 0.0005 |
The table is sorted according to d-values. UF, ultrafiltration; WWTP, wastewater treatment plant.
Pseudo-first-order rate constants (k-values)
It can be assumed that in a real use case, the water will only pass through the reactor once instead of being treated for 3 h. Therefore, removals after a single passage were calculated based on degradation constants. The concentration of each compound after having passed the UV reactor once (cUV) was calculated with Equation (3), where cm (t) is the concentration of the compound in the mixing vessel at the time when the water has passed the reactor once (where t is the residence time in the reactor, depending on the flow rate), Vmix is the volume of the mixing vessel, Q is the flow rate, and d is the degradation constant. The residence time of a particle in the reactor can be calculated by dividing the volume of the reactor ( approximately 400 mL) by the flow rate (see Figure 1). It is assumed that the concentration of each substance, cm (t), is constant in space in the parts colored in green. In the red part, the concentration has decreased by UV irradiation, and the concentration of each substance after UV irradiation, cUV, is constant in space.
The pseudo-first-order rate constants can be retrieved by plotting the calculated cUV values at different residence times from various experiments. Degradation constants (d) and concentrations in the mixing vessel (cm) can be calculated for each experimental setup with the help of the regression equations as determined in the DoE model.
RESULTS AND DISCUSSION
The oxidation by hydroxyl radicals leads to a nonspecific degradation of organic substances. Therefore, effects on the DOM represented by the DOC were characterized in the first step, before proceeding with the single substances OMP analysis and presentation of DoE results.
DOC characterization
Size-exclusion chromatograms with responses for organic carbon for selected experiments.
Size-exclusion chromatograms with responses for organic carbon for selected experiments.
The signals of fraction B (humic substances, detected at around 46 min) were higher for WWTP effluent compared to Mix and UF permeate. At the end of the experiments, humic substances were transformed and peak C (fulvic acids, detected at around 48 min) emerged. The transformation of humic substances into fulvic acids was less pronounced in the absence of H2O2, whereas UV radiation combined with H2O2 was effective in transforming humic substances. During the degradation of humic substances in the experiments due to irradiation and oxidation, compounds similar to humic substances but with lower molecular weight are formed, and these compounds are referred to as building blocks (Huber et al. 2011). The increase in fraction D (building blocks, detected at around 55 min) was likely due to UV radiation. The signals of building blocks at the end of the experiments were also high in the experiments with UV radiation and H2O2. When building blocks degrade further, compounds characterized by lower molecular weight and low ion density are formed (e.g., alcohols and sugars), and these compounds are referred to as low molecular weight neutrals (Huber et al. 2011). The signal of peak E (low molecular weight neutrals, detected at around 76 min) was increased through the addition of H2O2 and deployment of additional irradiation at 185 nm in some experiments (Figure 2(b)), suggesting the formation of transformation products. However, this effect is not visible in all experiments (Figure 2(g)), potentially due to the higher concentrations of substances (including particles) present in the unfiltered WWTP effluent. UV radiation alone (Figure 2(a), 2(d), and 2(f)) did not have a strong effect on fulvic acids, building blocks, and low molecular weight neutrals even though some transformation can be observed with a minor increase at around 70 min, most visible in Figure 2(a) and 2(f).
Excitation emission matrices (EEMs) of test solutions at 0, 40, and 180 min with signal intensities under two different operational conditions. The peaks can be identified as fulvic-like (A), humic-like (C), tryptophan-like (T), and SMP-like (S).
Excitation emission matrices (EEMs) of test solutions at 0, 40, and 180 min with signal intensities under two different operational conditions. The peaks can be identified as fulvic-like (A), humic-like (C), tryptophan-like (T), and SMP-like (S).
Degradation constants of all OMPs (experiment 14: 20 mg/L H2O2, WWTP effluent, dichromatic irradiation at 254 + 185 nm, flow rate 100 mL/s). Logarithmic regression lines are calculated with Equation (2). The dotted red line indicates 10% of the initial concentration. Substances are ordered according to Figure 6.
Degradation constants of all OMPs (experiment 14: 20 mg/L H2O2, WWTP effluent, dichromatic irradiation at 254 + 185 nm, flow rate 100 mL/s). Logarithmic regression lines are calculated with Equation (2). The dotted red line indicates 10% of the initial concentration. Substances are ordered according to Figure 6.
Calculated removals of OMPs after a single passage through the reactor (experiment 1: 20 mg/L H2O2, UF permeate, irradiation at 254 nm, flow rate 100 mL/s). Substances are ordered according to Figure 6.
Calculated removals of OMPs after a single passage through the reactor (experiment 1: 20 mg/L H2O2, UF permeate, irradiation at 254 nm, flow rate 100 mL/s). Substances are ordered according to Figure 6.
d-values from all experiments (black dots) and modelled distributions. The OMPs are grouped and ordered according to the mode value of the d-values (highest mode value in case of multimodal distribution). The maximum d-value calculated from regression equations are shown as red triangles.
d-values from all experiments (black dots) and modelled distributions. The OMPs are grouped and ordered according to the mode value of the d-values (highest mode value in case of multimodal distribution). The maximum d-value calculated from regression equations are shown as red triangles.
OMP degradation
Initial concentrations of OMP were between 0 and 10 μg/L prior to spiking with stock solution, except OXP (SI H). OXP was present in the highest concentrations in all WWTP effluent batches (20 ± 10 µg/L). Several of the 24 OMP were present in the effluent prior to spiking (see SI H). Notably, OXP, BTA, VSA, DCF, MEL, and DMBSA/XSA were present in high concentrations (median over 2 μg/L). SAC, TFMSA, CG, MPSA, PTSS, DPG, DIOTOG, HHTMP, AAMPS, ACE, and ATA were mainly added through spiking (median under 0.5 μg/L before spiking). The remaining substances (DZA, BETMAC, MAPMA, CBZ, SMX, BDMA, and PRI) were found at median concentrations between 0.5 and 2 μg/L before spiking. Concentration plots retrieved from experiment number 14 (WWTP effluent treated with UV light at 254 + 185 nm, 20 mg/L of H2O2, a flow rate of 100 mL/s) are shown in Figure 4 including the logarithmic regression lines based on the degradation constant (d).
Similar plots to those shown in Figure 4 for all DoE experiments can be found in SI I. Most of the substances showed a clear degradation from concentrations above 1 to below 0.1 μg/L. It is assumed that the spiking solution did not significantly alter the water matrix given the vast quantity of other wastewater constituents. As mentioned above, some substances (SAC, TFMSA, CG, MPSA, PTSS, DPG, DIOTOG, HHTMP, AAMPS, ACE, and ATA) were mainly added through spiking. For these substances, degradation constants might be slightly overestimated compared to the degradation in WWTP effluent with realistic concentrations.
After approximately 20 min of contact time, DCF was already degraded by 90%. Similar observations can be made in all experiments with UV light for ACE, DZA, and SMX. These are the most photo-susceptible compounds, as has been reported in several studies (Wols & Hofman-Caris 2012; Wols et al. 2013; Yu et al. 2015). BETMAC, DIOTOG, CBZ, MAPMA, DPG, VSA, DMBSA, and XSA, BDMA, PRI, MPSA, ATA, PTSS, AAMPS, and BTA took up to around 100 min for 90% degradation. Except for MAPMA, MPSA, ATA, and AAMPS, all OMP that were degraded by over 90% within 180 min of exposure contain cyclic structures.
In contrast, OXP, SAC, HHTMP, TFMSA, MEL, and CG did not reach 90% degradation after the full experiment duration of 180 min. CG even showed an increase in concentration of around 1 µg/L for most experiments. This effect might be due to water constituents, such as fertilizers or other nitrogen-containing substances, being transformed into CG (Rychter et al. 2019). Similarly, MEL, which is also a nitrogen-rich compound, showed increasing concentrations in some experiments.
Calculation of rate constants
Table 3 shows d- and k-values for DIOTOG as one example of OMP. The rate constants (k) of other OMPs can be found in SI J. From Table 3, we see that the highest rate constants were obtained in experiments with high H2O2 concentrations. All other factors played only a minor role. Within the range of natural pH variations (7.0–8.4), the pH value did not significantly influence d-values, as demonstrated by comparing d-values from experiments with identical setups but different pH values (experiment 12 with a pH of approximately 7 and experiment 13 with a pH of approximately 8). Dichromatic irradiation yielded higher reductions even without the use of H2O2. The usage of different wavelengths seems to have no influence when high H2O2 concentrations are present. Removals after a single passage through the reactor (calculated based on the degradation constants) were higher at a low flow rate (40 mL/s) due to the longer residence time in the reactor.
Removals of all substances after a single passage through the reactor are shown for experiment 1 in Figure 5. Substances susceptible to UV irradiation, such as DCF, are degraded substantially (up to around 45%). Substances with low rate constants are not significantly degraded in the same time period.
DoE analysis
For each of the 24 substances studied, 14 d-values corresponding to the 14 experimental conditions were calculated. Using DoE software, the d-values for each substance were modeled as a function of the input parameters flow rate, wavelength, water matrix, and H2O2 concentration, resulting in 24 linear regression equations (one for each substance; see SI K). In these simplified models, the regression equations do not account for interactions between the substances, but only between the input parameters.
The statistical errors of the linear regressions were significant for DCF, ACE, DMBSA/XSA, AAMPS, and BTA with p-values smaller than 0.05 (see Table K2 in SI K). These significant errors indicate non-linear behaviours of the d-values as a function of the input parameters and, consequently, reduced confidence in the predictive abilities of the corresponding linear regression equations. To account for non-linear relationships, a larger number of data points would be required for these substances.
The DoE regression equations can be used to identify combinations of process parameters linked to the highest degradation of the individual substances (Table 4). A Pareto chart of the standardized effects of all process parameters for each substance is presented in SI K.
Summary of DoE results for each OMP
OMP . | Assigned group . | Flow rate . | Wave-length . | Water matrix . | H2O2 . | Parameter with the largest standardized effect . |
---|---|---|---|---|---|---|
(mL/s) . | (nm) . | (mg/L) . | ||||
DCF* | 1 | 100 | 254* | UF permeate | 20 | Flow rate |
ACE* | 1 | 100 | 254* | UF permeate | 20* | Flow rate |
DZA | 1 | 100* | 254* | UF permeate* | 20* | Flow rate* |
SMX | 1 | 100 | 254 | UF permeate | 20 | Flow rate |
BETMAC | 2 | 100 | 254 + 185* | UF permeate* | 20 | H2O2 |
MAPMA | 2 | 40* | 254 + 185 | UF permeate* | 20 | H2O2 |
DIOTOG | 2 | 100 | 254 + 185* | UF permeate* | 20 | H2O2 |
CBZ | 2 | 100 | 254 + 185 | UF permeate | 20 | H2O2 |
DPG | 2 | 100 | 254 + 185 | UF permeate | 20 | H2O2 |
BTA* | 2 | 100 | 254* | UF permeate* | 20 | H2O2 |
VSA | 2 | 100 | 254 + 185 | UF permeate* | 20 | H2O2 |
BDMA | 2 | 100 | 254 + 185* | UF permeate* | 20 | H2O2 |
DMBSA/XSA* | 2 | 100* | 254 + 185* | UF permeate | 20 | H2O2 |
MPSA | 2 | 100 | 254 + 185 | UF permeate* | 20 | H2O2 |
AAMPS* | 2 | 100 | 254 + 185 | UF permeate | 20 | H2O2 |
PRI | 2 | 100 | 254 + 185 | UF permeate* | 20 | H2O2 |
ATA | 2 | 100 | 254* | UF permeate | 20 | H2O2 |
PTSS | 2 | 100 | 254 + 185* | UF permeate* | 20 | H2O2 |
HHTMP | 3 | 40 | 254 + 185 | WWTP effluent* | 0* | Wavelength |
SAC | 3 | 100 | 254* | UF permeate* | 20 | Flow rate · H2O2 |
OXP | 3 | 100 | 254 + 185 | WWTP effluent* | 20 | Wavelength |
TFMSA | 3 | 40* | 254 + 185* | UF permeate | 0* | Flow rate |
Water matrix | ||||||
MEL | 3 | 40 | 254 + 185 | UF permeate | 0 | H2O2 |
CG | 3 | 100* | 254 + 185* | WWTP effluent* | 0* | H2O2* |
OMP . | Assigned group . | Flow rate . | Wave-length . | Water matrix . | H2O2 . | Parameter with the largest standardized effect . |
---|---|---|---|---|---|---|
(mL/s) . | (nm) . | (mg/L) . | ||||
DCF* | 1 | 100 | 254* | UF permeate | 20 | Flow rate |
ACE* | 1 | 100 | 254* | UF permeate | 20* | Flow rate |
DZA | 1 | 100* | 254* | UF permeate* | 20* | Flow rate* |
SMX | 1 | 100 | 254 | UF permeate | 20 | Flow rate |
BETMAC | 2 | 100 | 254 + 185* | UF permeate* | 20 | H2O2 |
MAPMA | 2 | 40* | 254 + 185 | UF permeate* | 20 | H2O2 |
DIOTOG | 2 | 100 | 254 + 185* | UF permeate* | 20 | H2O2 |
CBZ | 2 | 100 | 254 + 185 | UF permeate | 20 | H2O2 |
DPG | 2 | 100 | 254 + 185 | UF permeate | 20 | H2O2 |
BTA* | 2 | 100 | 254* | UF permeate* | 20 | H2O2 |
VSA | 2 | 100 | 254 + 185 | UF permeate* | 20 | H2O2 |
BDMA | 2 | 100 | 254 + 185* | UF permeate* | 20 | H2O2 |
DMBSA/XSA* | 2 | 100* | 254 + 185* | UF permeate | 20 | H2O2 |
MPSA | 2 | 100 | 254 + 185 | UF permeate* | 20 | H2O2 |
AAMPS* | 2 | 100 | 254 + 185 | UF permeate | 20 | H2O2 |
PRI | 2 | 100 | 254 + 185 | UF permeate* | 20 | H2O2 |
ATA | 2 | 100 | 254* | UF permeate | 20 | H2O2 |
PTSS | 2 | 100 | 254 + 185* | UF permeate* | 20 | H2O2 |
HHTMP | 3 | 40 | 254 + 185 | WWTP effluent* | 0* | Wavelength |
SAC | 3 | 100 | 254* | UF permeate* | 20 | Flow rate · H2O2 |
OXP | 3 | 100 | 254 + 185 | WWTP effluent* | 20 | Wavelength |
TFMSA | 3 | 40* | 254 + 185* | UF permeate | 0* | Flow rate |
Water matrix | ||||||
MEL | 3 | 40 | 254 + 185 | UF permeate | 0 | H2O2 |
CG | 3 | 100* | 254 + 185* | WWTP effluent* | 0* | H2O2* |
Note. The OMPs are grouped and ordered according to the highest mode value of the distribution of d-values (Figure 6). The parameter combination (flow rate, wavelength, water matrix, and H2O2 concentration) yielding maximum d-values are shown. Substance marked with *: significant non-linear error (p < 0.05). parameter marked with *: parameter has a standardized effect below the significance line in the Pareto chart (SI K).
Based on the mode of the distributions of d-values,respectively, the highest mode value in the case of multimodal distributions, the substances were divided into three groups. Group 1 represents photodegradable substances with high d-values for which the most significant process parameter is the flow rate. Group 2 represents substances with intermediate d-values. These substances are only degraded in the presence of H2O2, and the H2O2 concentration is the most significant parameter. Finally, substances in group 3 are poorly removed and have low d-values.
The regression equations were used to calculate the maximum d-value for each substance. The input parameters were set to values, as defined in Table 4. For each substance, Figure 6 shows the distribution of d-values, as well as the maximum d-value retrieved from the regression equations. Generally, the maximum observed and modeled d-values were similar, indicating a good fit for the regression models. Exceptions are DZA and CG, where the maximum observed d-value exceeded the model prediction.
UVA254 as a surrogate parameter
UVA254 provides a characterization of the organics. UVA254 decreased during all experiments from a maximum value of around 30/m to a minimum value of around 14/m (overview of all experiments in SI L). Hence, absorbance at 254 nm decreased by up to 50%, which is similar to results from previous studies (Afonso-Olivares et al. 2016). Due to comparably higher persistence toward photolysis of UV-light-absorbing background organic matter, UV absorbance at 254 nm did not decrease as strongly as concentrations of OMPs studied for the photodegradable substances from group 1 (see SI L). In contrast, for several substances from groups 2 and 3, UVA254 abatement was higher than OMP abatement depending on the experimental conditions, in particular in the absence of H2O2.

Removal of three representative substances as a function of UVA254 abatement. Group 1: photosensitive substances; group 2: substances that are degraded by UV/H2O2; group 3: substances resistant to UV/H2O2 treatment. Black line, slope, and R2: linear regression between UVA254 abatement and OMP removal, capped at 90% OMP removal or at maximum UVA254 abatement.
Removal of three representative substances as a function of UVA254 abatement. Group 1: photosensitive substances; group 2: substances that are degraded by UV/H2O2; group 3: substances resistant to UV/H2O2 treatment. Black line, slope, and R2: linear regression between UVA254 abatement and OMP removal, capped at 90% OMP removal or at maximum UVA254 abatement.
Linear regressions between UVA254 abatement and OMP removal, capped at 90% OMP removal to capture the linear part of the relationships, were computed to evaluate the different behaviors of substances from the three groups (Figure 7). SMX (group 1) had the highest slopes and intercepts, followed by BETMAC (group 2) and SAC (group 3). High slopes indicate rapid reaction rates and positive vertical intercepts indicate that the elimination of OMP starts earlier than the elimination of UVA254 absorbance (Gerrity et al. 2012). R2 values were calculated to evaluate the goodness of fit. R2 ranged between 0.55 and 0.74. Overall, this relatively high fit indicates that it may be possible to establish substance-specific relationships between UVA254 abatement and OMP removal that are valid across a range of operational conditions and water matrices.
Practical implications
Overall, the results from this study highlight that UV/H2O2–AOP is a promising process for water reuse applications. The process was able to degrade a wide range of potentially PM OMP in several batches of WWTP effluent (with and without ultrafiltration), exhibiting realistic variability in scavenger substance activity. While the concentration of scavengers was not systematically varied in this study, the results show that degradation was generally higher in UF permeate (particle-free WWTP effluent). Other studies systematically assessed the effect of OH-radical scavengers (mainly carbonates and nitrite) and showed that degradation of OMP can be enhanced by removing scavenger substances, e.g., by stripping of CO2 at low pH values (Souza et al. 2014). Boczkaj & Fernandes (2017) and Liu et al. (2012) reported that degradation of most OMP is generally higher at lower pH values, making pH decrease a viable option in reuse systems. However, most WWTP effluents are near neutral (Rosario-Ortiz et al. 2010; Keen et al. 2012) and pH adjustment needs to be carefully evaluated with respect to the cost and complexity of acid addition. OMP degradation by UV and UV-H2O2 has been demonstrated to yield byproducts through nonselective hydroxylation, dechlorination, cyclization, and decarboxylation that can exhibit toxic properties to aquatic organisms and potentially also to humans (Huang et al. 2020). OMP degradation was generally higher (i) in UF permeate, (ii) with higher H2O2 concentrations, (iii) at lower flow rates, and (iv) using a dichromatic UV lamp. However, to make practical recommendations on the optimized implementation of UV/H2O2 for water reuse, it is necessary to define treatment targets for OMP removals. Only then will it be possible to compare the different options using relevant metrics such as treatment costs or energy requirements, e.g., the additional energy requirement for ultrafiltration of WWTP effluent vs. the lower energy requirement for UV/H2O2 post-treatment when ultrafiltration is used. Since water quality and quantity requirements vary by reuse application, the economic and ecological feasibility of UV/H2O2 post-treatment must be evaluated for each specific context (James et al. 2014). Once treatment requirements for OMPs have been defined for specific reuse applications, this study provides an example of an applicable approach to modeling the degradation of several OMPs as a function of major process parameters, thus allowing optimization of the UV/H2O2–AOP for different contexts.
CONCLUSIONS
This study provides degradation constants and pseudo-first-order rate constants for 24 OMPs for different configurations of UV or UV/H2O2–AOP post-treatment processes. DCF, ACE, DZA, and SMX were susceptible to UV irradiation alone. Several compounds not commonly studied, including BDMA, DIOTOG, HHTMP, MAPMA, XSA, showed degradability by the UV/H2O2–AOP. In contrast, MEL and CG were not susceptible to UV/H2O2–AOP and even showed increasing concentrations. A DoE approach was used to model the degradation of OMPs as a function of the process parameters, enabling a substance-specific optimization of the UV/H2O2–AOP post-treatment. The 24 OMPs were divided into three groups according to the modal values of the degradation constants. This study further demonstrated that it is possible to establish substance-specific relationships between UVA254 abatement and OMP removal across operational conditions and water matrices, underscoring the potential of using UVA254 abatement as a proxy to predict OMP degradation. Overall, the study highlights that UV/H2O2–AOP is an interesting treatment process for water reuse due to its capability to degrade a wide range of potentially mobile and persistent OMPs. However, to achieve substantial OMP removal, it is necessary to recirculate the water or implement several UV/H2O2 reactors in series.
ACKNOWLEDGEMENTS
This investigation was partly supported by the German Federal Ministry of Education and Research (BMBF) within the project PU2R (contract 02WV1564A). We thank Silke Pabst, Andrea Steuer, Fanny Kohn-Eberle, Leon Saal, and Pia Schumann for their support, as well as the Berliner Wasserbetriebe (BWB) for providing WWTP effluent.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.